What Drives Electrician Headcount in a Data Hall?

At the MDN Data Center in Midlothian, Texas, Cummings Electrical coordinated a peak of 430 electricians. The building was 60 MW and 300,000 square feet, and the contractor's project case study lists four 1200A, 25kV main-tie-main switchgear lineups, 48 distribution skids at 4000A and 480V/415V, and 48 generators rated 2750 kW each. The job ran from October 2018 to May 2020 and executed more than 130 MOPs.

Divide 430 by 60 and the peak works out to roughly 7.2 electricians per megawatt.

Keep that number in mind when someone proposes staffing a data hall off a per-MW ratio or the last hall's crew count. Headcount is the last number you calculate, after quantities, labor units and the room you can actually work in.

Per-megawatt ratios spread too wide to staff a hall

The most careful public benchmark I have found is a 2025 report by Angela Ryu and Shon R. Hiatt of the USC Zage Business of Energy Initiative, published by the Hamm Institute. Its summary puts construction at 0.7 to 2.0 workers per MW, measured as peak direct on-site construction headcount. Its detailed benchmark table puts campuses of 100 MW or more built with modular and prefab methods at 0.6 to 0.8, and sites under 20 MW or custom builds at 1.5 to 2.0.

Their own benchmark projects scatter from 1.3 jobs per MW at a campus in Abilene, Texas to 3.6 at Vantage Frontier. The authors are candid about why. "No single source provides a complete, consistent breakdown of direct vs. indirect employment, construction vs. operations staffing, or MW-specific headcounts," they wrote.

Midlothian's electricians alone came in at several times the all-trade range. Different era, different design. Still, a ratio that can miss by that much is good for a sanity check, and I would not build a crew plan on it.

Labor units start the count, and supervision sits outside them

Start from the takeoff. Take the released quantity for each task in the hall, subtract what has been installed and accepted, and multiply by a labor unit that fits the conditions. The NECA Manual of Labor Units, 2025-2027 edition, gives each item normal, difficult and very difficult columns, and this edition adds more than 150 new items, expands coverage of UPS equipment and batteries, and revises the bus duct descriptions.

Read the fine print on that page. NECA says its units cover normal material handling, drawing study, measurement and layout, installation and normal non-productive labor, and no supervision of any type. Most electrical contractors, the page says, think "the labor for supervision is best estimated as a separate cost item."

So the foremen go on their own line. QA/QC, controls and test support usually get their own lines too.

The units themselves still need adjusting. George Hague, then president of ConEst, wrote in a 2000 Electrical Contractor article that "a labor unit is not absolute; it is a benchmark, or starting point," and gave mounting-height adders of 5 percent over 10 feet and 10 percent at 16 to 20 feet, with a lift or scaffold allowance above 20 feet.

Divide the remaining hours by the effective installer hours in the window (observed hours from comparable work, never paid shift hours) and you have a workload number.

The workfront sets the ceiling

The room has a limit. Awad S. Hanna of UW-Madison measured this for electrical contractors, and in a 2004 Electrical Contractor article on trade stacking he wrote that "there is 0 percent lost efficiency at a worker density of 191 square feet per worker and a 50 percent loss of efficiency at a worker density of 90 square feet per worker." The UW-Madison research put the minimum at 200 to 250 square feet per worker for full productivity, depending on the complexity of the work. Worker density counts every trade in the net working area, and the model applies only to projects of 2,000 to 150,000 man-hours.

It is older research, and it tends to match what happens in the last months of a data hall.

In a NECA member survey reported in the same article, contractors placed nearly half of their stacking in the final quartile of construction. In practice, walk the released zone, measure the floor left after lift aisles and staging, divide by roughly 200 square feet and subtract the other trades booked there that week. If the workload number is higher than that ceiling, you have a constraint, and more people will not fix it. Open another zone, finish the material kits, change the sequence or move the date.

Overtime beat overmanning in the electrical data

Hanna came back to this in a 2011 ELECTRI International measured-mile study of electrical projects built from 2004 to 2008. Of the 400 contractors surveyed, 51 answered. "Despite productivity losses related to overtime, placing an electrical crew on overtime schedule is more efficient than overmanning," the report says. Extra labor on site makes access to tools, materials and space much harder to maintain, it adds.

Overtime has a price too. On a 5x10 schedule the productivity factor fell to 0.84 by week 4 and about 0.71 in weeks 9 through 11. One tracked 6x10 job, a 36-story Florida condominium, bottomed out at 55 percent in its final week, and the report itself warns that overtime factors from NECA and similar studies "should be used as a guide and not as absolute qualification." For a second shift, Hanna's 2004 article suggested asking the first-shift foreman to stay an hour late and the second-shift foreman to come an hour early, or giving the second shift completely different tasks, locations and materials. The Friday version of that trade-off is laid out in Overtime or flex crew: run the math before Friday.

In HKA's 120 MW scenario, the sub staffed 60 percent

Consultants at HKA laid out a hypothetical 120 MW colocation project in Texas, built from public information and the experience of HKA staff. In it, the electrical subcontractor forecast a peak of about 150 electricians for medium-voltage work, power distribution, standby generation and UPS installation, all running at once. Despite sustained overtime, premium pay and staggered shifts, the sub consistently staffed about 60 percent of that. The authors describe what followed, including improper terminations, noncompliant grounding, undocumented field changes and repeated punch and rework cycles, until commissioning could not proceed as planned and turnover milestones slipped.

How fast the job ramps matters too. Carmen Voso, business agent for IBEW Local 24, was looking at the Quantum Frederick campus in Maryland when he told the union's Electrical Worker what he hoped for. "Our hope is that they ramp up slowly and it isn't 'We need 1,000 in six weeks, then another 1,000,'" Voso says. The curve by phase is covered in the data center crew curve from underground to commissioning.

Skill mix changes the answer

Andrew Mason, a general foreman at Power Solutions, runs a Loudoun County, Virginia crew of more than 250 electricians. In a Data Center Frontier piece sponsored by IBEW Local 26, he puts his apprentices at "60-70 at any given time," first through fifth year, or roughly one in four. "On any given site you can see a large portion of the electrical industry from medium voltage systems, fire alarms, to controls, everything is covered," Mason says.

Apprentice ratios are also set in writing. Federal apprenticeship standards at 29 CFR 29.5(b)(7) require a numeric ratio of apprentices to journeyworkers consistent with proper supervision, training and safety. And a green hire costs production early. Scotty Wristen, owner of WE Electric in Abilene, lost five workers to data center jobs paying $35 an hour against his $20. He told The Texas Tribune that new apprentices usually need four or five months to get up to speed, and the small mistakes along the way cost time and money. "It's fixable," he says.

Energization and testing shrink the room

As parts of a hall move toward energization, the usable workfront gets smaller. Under 29 CFR 1926.416(a)(3), the employer must determine before work begins whether any energized circuit is located where people, tools or machines could contact it, and must tell employees where it is, what the hazards are and what protection is required. Section 1926.417(b) requires tags at every point where a deenergized circuit can be energized.

Testing needs its own crew line. Robert Hill, regional director of acceptance testing at Vector Power, wrote in NETA World that "the ideal scenario is to perform acceptance testing once equipment is installed in its final position and before it is energized," and that "no single technician can address every aspect of a system of this size and complexity."

The people to fill those lines are scarce

In EC&M's 2026 Top 50 report, 83 percent of responding Top 50 firms reported labor shortages, nine points more than the year before. The AGC and NCCER survey released September 3 drew 1,830 respondents, and among firms that did data center work in the past year, 58 percent said those projects increased competition for skilled workers.

Sources

1. MDN Data Center, Midlothian, TX, Cummings Electrical, project 2018 to 2020, accessed September 29, 2026.
2. Data Center Employment Forecast Analysis, Angela Ryu and Shon R. Hiatt, Hamm Institute, November 2025.
3. NECA Manual of Labor Units (MLU), 2025-2027 Edition, National Electrical Contractors Association, accessed September 29, 2026.
4. Understanding Labor Units, George Hague, Electrical Contractor, January 15, 2000.
5. The Stacking of Trades, Awad S. Hanna, Electrical Contractor, August 15, 2004.
6. Impact of Overtime on Electrical Labor Productivity: A Measured Mile Approach, Awad S. Hanna, ELECTRI International, 2011.
7. Data center construction in the age of AI: Emerging risks, delays, and disputes, Jorge Berrios, Scott Hollingsworth and Daniel Kwon, HKA, September 4, 2026.
8. The Data Center Surge: A New Generation of IBEW Jobs, The Electrical Worker, IBEW, May 1, 2025.
9. Views from the Ground: Electrical Contractors Building Data Centers Weigh In, David Pala, Data Center Frontier (sponsored by IBEW Local 26), November 13, 2024.
10. 29 CFR 29.5, Standards of apprenticeship, Legal Information Institute, Cornell Law School, accessed September 29, 2026.
11. Data centers, Texas home builders compete for electricians, Sneha Dey, The Texas Tribune, April 28, 2026.
12. 29 CFR 1926.416, General requirements, Occupational Safety and Health Administration, accessed September 29, 2026.
13. 29 CFR 1926.417, Lockout and tagging of circuits, Occupational Safety and Health Administration, accessed September 29, 2026.
14. ANSI/NETA Standards in Data Center Acceptance Testing, Robert Hill, NETA World Journal, February 7, 2026.
15. Data Center Buildout Fuels Revenue Blowout: EC&M's 2026 Top 50 Electrical Contractors Special Report, Tom Zind, EC&M, September 15, 2026.
16. Construction Workforce Shortages Remain Acute Despite 'Soft' Market Conditions As Data Centers Strain Labor Supply, Survey Finds, Associated General Contractors of America, September 3, 2026.

Featured image: MTA Capital Construction, via Wikimedia Commons (CC BY 2.0).